Type 1 (insulin-dependent) diabetes mellitus is associated with HLA DR and DQ factors, but the primary risk alleles are difficult to identify because recombination events are rare in the DQ-DR region. The risk of HLA genotypes for type 1 diabetes was therefore studied in more than 420 incident new onset, population-based type 1 diabetes children and 340 age, sex and geographically matched controls from Sweden. A stepwise approach was used to analyse risk by relative and absolute risks, stratification analysis and the predispositional allele test. The strongest relative and absolute risks were observed for DQB1*02-DQA1*0501/DQB1*0302-DQA1*0301 heterozygotes (AR 1/46, P < 0.001) or the simultaneous presence of both DRB1*03 and DQB1*0302 (AR 1/52, P < 0.001). Stratification analysis showed that DQB1*0302 was more frequent among DRB1*04 patients than DRB1*04 controls (P < 0.001), while DRB1*03 was more frequent among both DQA1*0501 (P < 0.001) and DQB1*02 (P < 0.001) patients than respective controls. The predispositional allele test indicated that DRB1*03 (P < 0.001) would be the predominant risk factor on the DRB1*03-DQA1*0501-DQB1*02 haplotype. In contrast, although DQB1*0302 (P < 0.001) would be the predominant risk factor on the DRB1*04-DQA1*0301-DQB1*0302 haplotype, the predispositional allele test also showed that DRB1*0401, but no other DRB1*04 subtype, had an additive risk to that of DQB1*0302 (P < 0.002). It is concluded that the association between type 1 diabetes and HLA is due to a complex interaction between DR and DQ since (1) DRB1*03 was more strongly associated with the disease than DQA1*0501-DQB1*02 and (2) DRB1*0401 had an additive effect to DQB1*0302. The data from this population-based investigation suggest an independent role of DR in the risk of developing type 1 diabetes, perhaps by providing diseases-promoting transcomplementation molecules.